SW-actors: accelerating the Smith-Waterman algorithm via actors.

Abstract

Motivation: The Smith-Waterman (SW) algorithm is widely regarded as the gold standard for local sequence alignment. However, its time complexity in a serial implementation limits its practicality for large datasets. In this article, we introduce SW-actors, a parallel implementation of the SW algorithm that leverages the actor model of concurrent computation to optimize resource utilization by efficiently scheduling and managing independent alignment tasks across processors at both the interalignment and intraalignment levels.

Results: SW-actors is compared with the state-of-the-art implementations Parasail, SeqAn, and SWIPE using four datasets of varying sequence lengths ranging from 85 to 74778 nucleotides. In terms of wall-clock time, SW-actors is 1.33 $\times$ , 2.00 $\times$ , 2.49 $\times$ , and 1.94 $\times$ faster than the next best implementation for the different datasets. SW-actors is up to 22 $\times$ faster than serial on 40 cores. The speedup is consistent for larger datasets and hence offers significant advantages for medium- to large-scale alignments.

Availability and implementation: The SW-actors source code and underlying data are available at https://git.cs.usask.ca/numerical_simulations_lab/actors/papers/sw-actors.